FaceForensics++: 7,010 Videos for Deepfake Detection

7,010 MP4 video files, including 6,000 deepfakes and 1,000 real videos, were downloaded using original scripts. The dataset was uploaded by user 'bitmind' to Hugging Face and last updated on October 29, 2025. It contains deepfakes generated by six different methods: DeepFakeDetection, Deepfakes, Face2Face, FaceShifter, FaceSwap, and NeuralTextures.

Use Cases

• Training deepfake detection models based on the six distinct manipulation methods described.
• Benchmarking the performance of video forensic algorithms against a known set of 6,000 synthetic videos.
• Analyzing the visual artifacts and characteristics of different face-swapping and neural texture techniques.
• Developing general-purpose video classification models to distinguish real from synthetic content.

Strengths

• Contains 7,010 total video files, providing a substantial collection for model training.
• Includes 1,000 real videos as a baseline for comparison against 6,000 synthetic examples.
• Covers six distinct deepfake generation methods, allowing for method-specific analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and specific file formats beyond MP4 are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Hugging Face, uploaded by user 'bitmind'.

Collection Method

Downloaded using the original FaceForensics++ dataset scripts.

Time Range

null

Freshness

Last updated 2025-10-29 16:10:55; freshness should be verified.

Geography

null